Intelligent warehouse management system and method for automobile parts based on Internet of Things

Abstract

The invention belongs to the technical field of big data processing, and discloses an intelligent storage management system and method for automobile parts based on the Internet of Things. Comprising the steps of collecting attribute data of automobile parts, performing part labeling and data cleaning in sequence, and constructing a part basic database; original sensing data is collected and associated with the accessory basic database, a multi-source storage data set is obtained, accessory partitioning is carried out, and optimal storage layout information is generated; acquiring real-time position information of the accessories, and outputting an efficient work optimization path in combination with the optimal storage layout information; accessory warehouse-in and warehouse-out records are collected in real time, state prediction is carried out, and referable inventory information is output. Performing quality detection on the accessory basic database based on the referable inventory information, and outputting accessory evaluation data; collecting behavior data of a worker and carrying out behavior pattern analysis to generate behavior evaluation data; a storage strategy is constructed, and the storage strategy is sent to a preset automobile part management system; and high-efficiency and high-precision warehouse management is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of big data processing, and more particularly, to an intelligent automobile parts warehouse management system and method based on the Internet of Things. BACKGROUND

[0002] With the continuous development of intelligent manufacturing, industrial internet and warehouse automation technology, the automobile manufacturing field has higher requirements for the intelligentization of parts warehouse management; in order to meet the increasingly complex parts circulation and multi-scene application requirements, most enterprises have gradually begun to use intelligent warehouse management systems for warehouse management to improve warehouse operation and management efficiency and intelligent level, but there are still certain challenges in the prior art field.

[0003] In the warehouse management system in the prior art field, the data sources of automobile parts are scattered, and the types are various and prone to state dynamic changes, some data may only record the code or model, lack of attribute information of the parts and the correlation information between the parts, which leads to the inability to support the normal progress of the work; secondly, although the warehouse management system used in the prior art field is equipped with a positioning device, it cannot semantically associate real-time location information with automobile parts data, leading to disordered partitioning of parts and prone to mispicking or scheduling conflicts, for example, parts of the same vehicle model are allocated to different areas due to data classification errors, resulting in a decrease in the work efficiency of the workers picking; at the same time, the workers in the existing warehouse management system rely on experience to develop work paths, lack of path guidance and real-time evaluation, and prone to work path congestion problems due to repeated actions and path overlap, which seriously affects the work efficiency; in addition, the existing technology field uses fixed thresholds for judgment in inventory management, lacks dynamic adaptability, and cannot timely judge inventory abnormalities; and in the aspect of parts quality detection, it relies on manual sampling inspection, lacks effective analysis of environmental factors and parts life cycle data, and is difficult to achieve high-precision quality grading judgment; the behavior mode evaluation of the workers in the prior art field is also not universally configured, lacks analysis means for the operation deviation and behavior safety of the workers, and leads to abnormal behavior phenomena such as path deviation or task execution sequence disorder in the work process, which cannot be timely warned and optimized, resulting in a decrease in the overall work efficiency.

[0004] In view of this, the present application provides an intelligent automobile parts warehouse management system and method based on the Internet of Things to solve the above problems. SUMMARY

[0005] In order to overcome the above-mentioned defects of the prior art, in order to achieve the above-mentioned purpose, the present application provides the following technical scheme: an intelligent automobile parts warehouse management method based on the Internet of Things, comprising:

[0006] S1. Collect automobile parts attribute data and perform parts labeling to obtain a labeled parts dataset; perform data cleaning on the standard parts dataset to obtain an accurate parts dataset; and construct a parts basic database based on the accurate parts dataset;

[0007] S2. Collect original perception data and associate the parts basic database to obtain a multi-source storage dataset; perform parts partitioning on the multi-source storage dataset to generate optimal warehouse layout information;

[0008] S3. Obtain parts real-time location information, perform path optimization on parts picking tasks to be executed based on the optimal warehouse layout information and the parts real-time location information, and output an efficient work optimization path;

[0009] S4. Collect parts in-out warehouse records in real time, perform state prediction on the parts in-out warehouse records, and output referenceable inventory information;

[0010] S5. Perform quality detection on the parts basic database based on the referenceable inventory information, and output parts evaluation data;

[0011] S6. Collect behavior data of workers, perform behavior pattern analysis on the behavior data, and generate behavior evaluation data;

[0012] S7. Construct a warehouse strategy based on the behavior evaluation data, the parts evaluation data, the referenceable inventory information, and the efficient work optimization path, and send the warehouse strategy to a preset automobile parts management system.

[0013] Further, the way of collecting original perception data and associating the parts basic database comprises:

[0014] Collect original perception data and perform feature extraction on the original perception data to obtain original perception features; construct a parts semantic dictionary based on a preset automobile parts parameter database; extract parts labels of the parts basic database, group the parts basic database based on the parts labels, and establish a unique correspondence between each parts label and the corresponding group of parts basic data; perform semantic matching between the original perception features and each group of parts basic data and parts labels based on the parts semantic dictionary to obtain a multi-source storage dataset.

[0015] Further, the way of performing parts partitioning on the multi-source storage dataset comprises:

[0016] Perform feature extraction on the multi-source storage dataset to obtain structured semantic features; construct a feature vector space and map all structured semantic features to the feature vector space to obtain parts feature vectors; perform parts feature classification on all parts feature vectors to obtain parts feature classification results; and divide parts with the same category into the same warehouse area based on the parts feature classification results to obtain an initial parts partitioning set;

[0017] Real-time extraction of accessory warehouse layout data, construction of partition allocation filtering model based on accessory warehouse layout data; construction of optimal filtering function, and the optimal filtering function as the loss function of the partition allocation filtering model; using the trained partition allocation filtering model to perform partition filtering on the initial accessory partition set to obtain the optimal accessory partition set; mapping the optimal accessory partition set to the specific warehouse shelf coordinates to obtain the optimal warehouse layout information.

[0018] Further, the way of classifying all accessory feature vectors includes:

[0019] Based on all accessory feature vectors, construct an accessory topology network structure; calculate the feature similarity of any two accessory feature vectors, and weight the connection edge between the two accessory feature vectors using the feature similarity;

[0020] Classify all accessory nodes in the accessory topology network structure, set a similarity threshold, and perform coarse classification based on the feature similarity and the similarity threshold, filter the accessory node pairs with feature similarity greater than the similarity threshold, obtain the initial classification node set, and construct an accessory topology network subgraph based on the initial classification set; calculate the connection strength between the subgraph accessory nodes in any one accessory topology network subgraph, and perform detailed classification based on the connection strength to obtain the fine classification node set; integrate all fine classification node sets to obtain the accessory feature classification result.

[0021] Further, the way of optimizing the path of the accessory picking task to be performed includes:

[0022] Collect historical path information, perform path modeling on the historical path information to obtain a working path model; construct a warehouse state space, perform path reasoning on the accessory picking task to be performed based on the warehouse state space and the working path model, obtain feasible paths, integrate all feasible paths to generate a feasible path set; construct a path optimization objective function, perform path planning on the feasible path set based on the path optimization objective function to generate a feasible path score; set a path score threshold, filter the optimal feasible path based on the feasible path score and the path score threshold, which is the efficient work optimization path;

[0023] Deploy the preset path guidance unit, send the efficient work optimization path to the path guidance unit, and connect the path guidance unit with the wearable device; use the path guidance unit to visually display the efficient work optimization path, generate a virtual work path scene and send it to the wearable device; use the wearable device to monitor the work path trajectory in real time, and use the path optimization model to generate the latest efficient work optimization path based on the work path trajectory to adjust the work path trajectory.

[0024] Further, the manner of predicting the state of the accessory warehouse entry and exit record comprises:

[0025] Extracting accessory attribute information of the accessory warehouse entry and exit record corresponding to the automobile accessory; time sorting the accessory warehouse entry and exit record based on the timestamp of the accessory warehouse entry and exit record to obtain an entry and exit time sequence; data fusion of the accessory attribute information and the entry and exit time sequence to obtain an entry and exit feature time sequence;

[0026] Constructing a dynamic sliding window, traversing the entry and exit feature time sequence using the dynamic sliding window; constructing an accessory state evaluation system; predicting the state of each entry and exit feature time sequence segment in the dynamic sliding window, outputting a state evaluation score group based on the accessory state evaluation system; designing an accessory inventory state classification table based on a preset automobile accessory parameter database, comparing the state evaluation score group with the accessory inventory state classification table, and outputting inventory state prediction data of the entry and exit feature time sequence segment in the single dynamic sliding window; by continuously adjusting the size of the dynamic sliding window, the state prediction data of the entry and exit feature time sequence segment in each dynamic sliding window is weighted and fused to obtain referenceable inventory information.

[0027] Further, the manner of detecting the quality of the accessory database comprises:

[0028] Extracting accessory life cycle data of the accessory database, and performing feature fusion of the accessory life cycle data and the referenceable inventory information of the corresponding automobile accessory to obtain a quality detection feature vector; mapping each quality detection feature vector to an accessory life cycle sequence constructed based on the timestamp to obtain quality detection trajectory data;

[0029] Constructing an accessory quality rule database and taking the accessory quality rule database as a constraint condition; performing quality detection on the quality detection trajectory data based on the accessory quality rule database to output a quality evaluation score and an abnormal state label; setting and dynamically adjusting a quality score classification threshold, determining the quality classification corresponding to each automobile accessory based on the quality evaluation score and the quality score classification threshold; performing risk control on the abnormal state label and the quality classification corresponding to each automobile accessory, generating a risk management recommendation instruction for the automobile accessory in combination with the accessory quality rule database; and integrating the quality classification, the abnormal state label, and the risk management recommendation instruction to obtain accessory evaluation data.

[0030] Further, the manner of analyzing the behavior data comprises:

[0031] The behavior data is associated with the accessory warehouse in and out record to obtain multi-dimensional behavior data, and the worker ID tag corresponding to the behavior data is obtained, and the ID tag is used to mark the multi-dimensional behavior data; entity extraction is performed on the multi-dimensional behavior data to obtain behavior entities; a topological graph structure is constructed based on the behavior entities, each behavior entity is taken as a node of a work behavior graph, a logical relationship between any two behavior entities is extracted, and the two behavior entities are semantically connected by using the logical relationship to obtain the work behavior graph; the behavior mode subgraph of the corresponding worker in the work behavior graph is extracted based on the ID tag, and the behavior path of the corresponding worker is obtained by traversing each node and edge in the behavior mode subgraph;

[0032] The historical high-evaluation behavior data is collected, and a standard behavior path template is constructed based on the historical high-evaluation behavior data; the standard behavior path template is compared with the behavior path, and a behavior deviation feature vector is obtained by calculating the deviation degree; behavior evaluation information is obtained by performing behavior evaluation on the behavior deviation feature vector; the behavior evaluation information is bound with the ID tag of the corresponding worker, and all behavior evaluation information is integrated to obtain behavior evaluation data; and high-evaluation behavior data is obtained by performing efficient behavior screening on the behavior evaluation data, and the standard behavior path template is dynamically updated by using the high-evaluation behavior data.

[0033] Further, the method of constructing a warehouse strategy based on behavior evaluation data, accessory evaluation data, referenceable inventory information and efficient work optimization path comprises:

[0034] The behavior evaluation data, accessory evaluation data, referenceable inventory information and efficient work optimization path are respectively subjected to feature extraction to obtain behavior features, accessory quality features, inventory information features and path features; the behavior features, accessory quality features, inventory information features and path features are subjected to feature fusion to obtain a multi-dimensional joint feature vector; a judgment index is set for each dimension of the multi-dimensional joint feature vector, and the judgment index is matched with a preset strategy rule database, and the multi-dimensional joint feature vector corresponding to all judgment indexes that can be matched with the strategy rule database is taken as the warehouse strategy.

[0035] A vehicle accessory intelligent warehouse management system based on the Internet of Things is used to implement a vehicle accessory intelligent warehouse management method based on the Internet of Things, and is characterized by comprising:

[0036] A data acquisition module is used to acquire vehicle accessory attribute data and perform accessory labeling to obtain a labeled accessory dataset; a standard accessory dataset is subjected to data cleaning to obtain a precise accessory dataset; and an accessory basic database is constructed based on the precise accessory dataset;

[0037] The accessory partition module is used for collecting original perception data and associating with an accessory database, obtaining a multi-source storage data set; the multi-source storage data set is partitioned to generate optimal storage layout information;

[0038] The path optimization module is used for obtaining real-time position information of accessories, optimizing the path of the accessory picking task to be performed based on the optimal storage layout information and the real-time position information of the accessories, and outputting an efficient work optimization path;

[0039] The inventory prediction module is used for collecting real-time accessory warehouse entry and exit records, predicting the state of the accessory warehouse entry and exit records, and outputting referenceable inventory information;

[0040] The quality detection module is used for detecting the quality of the accessory database based on the referenceable inventory information, and outputting accessory evaluation data;

[0041] The behavior analysis module is used for collecting behavior data of workers, analyzing the behavior pattern of the behavior data, and generating behavior evaluation data;

[0042] The strategy generation module is used for constructing a warehouse strategy based on the behavior evaluation data, the accessory evaluation data, the referenceable inventory information and the efficient work optimization path, and sending the warehouse strategy to a preset automobile accessory management system; each module is connected through wired and / or wireless means.

[0043] The technical effects and advantages of the automobile accessory intelligent warehouse management system and method based on the Internet of Things are as follows:

[0044] An automobile accessory intelligent warehouse management method and system based on the Internet of Things are realized from the aspects of data collection, perception fusion, partition optimization, path planning, inventory prediction, quality monitoring, behavior analysis and strategy generation, which significantly improves the overall operation efficiency and decision-making ability of the warehouse system. Compared with the traditional method relying on manual experience and static strategy, the method can fully integrate accessory data, Internet of Things perception data and behavior data, establish semantic mapping and state perception of multi-source information, improve accessory storage layout and work efficiency through intelligent partition and path optimization, enhance risk prevention and control capability through prediction and detection of inventory state and accessory quality, and indirectly optimize the work level of workers by analyzing their behavior data. Therefore, the automobile accessory intelligent warehouse management method and system based on the Internet of Things has the technical advantages of high intelligence, strong robustness and good self-adjusting ability, and can be widely applied to warehouse management scenes in the fields of intelligent manufacturing and automobile service, and has good practical value and promotion prospect. BRIEF DESCRIPTION OF DRAWINGS

[0045] Figure 1 A schematic diagram of the automobile accessory intelligent warehouse management method based on the Internet of Things is shown in the figure.

[0046] Figure 2 Figure 1 is a schematic diagram of an automobile accessory intelligent warehouse management system based on the Internet of Things according to the present application. DETAILED DESCRIPTION

[0047] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0048] Embodiment 1

[0049] Please refer to Figure 1 The automobile accessory intelligent warehouse management method based on the Internet of Things described in the present embodiment comprises:

[0050] S1. Collecting automobile accessory attribute data and performing accessory labeling to obtain a labeled accessory dataset; performing data cleaning on the standard accessory dataset to obtain a precise accessory dataset; and constructing an accessory basic database based on the precise accessory dataset;

[0051] S2. Collecting original perception data and associating the accessory basic database to obtain a multi-source storage dataset; performing accessory partitioning on the multi-source storage dataset to generate optimal warehouse layout information;

[0052] S3. Obtaining accessory real-time position information, performing path optimization on accessory picking tasks to be executed based on the optimal warehouse layout information and the accessory real-time position information, and outputting an efficient work optimization path;

[0053] S4. Real-time collection of accessory warehouse entry and exit records, state prediction of the accessory warehouse entry and exit records, and output of referenceable inventory information;

[0054] S5. Quality detection of the accessory basic database based on the referenceable inventory information, and output of accessory evaluation data;

[0055] S6. Collection of worker behavior data, behavior pattern analysis of the behavior data, and generation of behavior evaluation data;

[0056] S7. Construction of a warehouse strategy based on the behavior evaluation data, the accessory evaluation data, the referenceable inventory information, and the efficient work optimization path, and sending of the warehouse strategy to a preset automobile accessory management system.

[0057] The attribute data of the automobile parts is labeled by using an RFID electronic tag to obtain a labeled parts dataset; the data cleaning method for the labeled parts dataset includes outlier processing, missing value filling, and data denoising to obtain an accurate parts dataset; the missing fields are identified and filled based on existing knowledge or expert experience, an outlier detection model based on an isolation forest model is constructed for outlier processing to remove irrelevant fields or redundant information in the data, and data denoising is achieved; it should be noted that the parts basic database includes parts coding, specifications, compatible vehicle models, suppliers, and life cycle data.

[0058] The way of collecting original perception data and associating the parts basic database includes:

[0059] The original perception data is collected by a multi-level Internet of Things perception model deployed in the warehouse area, including parts location, parts surrounding environment, and parts quantity; the original perception data is feature extracted to obtain location features, environmental parameter features, and parts state features, etc., to form original perception features in a unified format, facilitating subsequent semantic association; a parts semantic dictionary is constructed based on a preset automobile parts parameter database, wherein the automobile parts parameter database is constructed based on existing field materials and expert experience, including attribute information related to automobile parts; the parts semantic dictionary is constructed by extracting semantic relationships from the automobile parts parameter database, including, for example, classification labels of automobile parts, parts coding, compatible vehicle models, and other basic information, and also including semantic relationship information such as parts usage; the parts labels of the parts basic database are extracted, the parts basic database is grouped based on the parts labels, and a unique correspondence between each parts label and the corresponding group of parts basic data is established, wherein the parts labels include the function labels and the parts coding of each automobile part, the grouping is first based on the parts coding, and then the grouping of the parts basic information in each group is based on the function labels; the grouping is first based on the parts coding to form large groups, and then based on the same function to form small groups, facilitating subsequent semantic matching to quickly match parts basic data containing the same function; the original perception features are semantically matched with each group of parts basic data and parts labels based on the parts semantic dictionary to obtain a multi-source storage dataset, the semantic structure mapping between the original perception features and the parts basic data and the corresponding parts labels is established by calling the parts semantic dictionary, the parts basic data corresponding to each group of parts labels is matched with the usage, function, category, and location of the corresponding automobile parts in the parts semantic dictionary, and the association between the original perception data and the parts basic database is achieved.

[0060] The way of partitioning the parts based on the multi-source storage dataset includes:

[0061] The structured semantic features are obtained by feature extraction on the multi-source storage data set. Since the multi-source storage data set contains semantic relationships between automobile parts and other data, semantic feature extraction is performed to obtain semantic features including attribute information and functional use, i.e., structured semantic features. A feature vector space is constructed, and all structured semantic features are mapped to the feature vector space to obtain part feature vectors. All structured semantic features are converted into vector representations through a preset vector encoding method and mapped to the same feature vector space. The accuracy and dimension of all features are unified, and the dimensional difference is eliminated, laying a foundation for subsequent data processing. The part feature classification is performed on all part feature vectors to obtain part feature classification results. In this embodiment, the part feature classification is implemented by constructing a graph structure of all part feature vectors and then performing hierarchical clustering. Based on the part feature classification results, parts with the same category are divided into the same storage area to obtain an initial part partition set. Parts with the same feature category are merged into the same group, and a storage area is assigned to the parts in each group. In this embodiment, only the classification results are considered, and the region is divided based on the classification results to obtain the initial part partition set.

[0062] Real-time extraction of part storage layout data, construction of a partition allocation screening model based on part storage layout data, wherein the part storage layout data includes part correlation, turnover rate, and storage conditions, etc. The partition allocation screening model is based on a multi-objective particle swarm optimization algorithm. An optimal screening function is constructed, and the optimal screening function is used as the loss function of the partition allocation screening model. The optimal screening function is the optimization objective function of the multi-objective particle swarm optimization algorithm, which is used as the loss function of the partition allocation screening model to guide the model to find the optimal solution. The calculation formula of the optimal screening function is: L0=α1×FR+α2×CB+α3×TP; wherein L0 represents the function value of the optimal screening function; FR represents the idle area; CB represents the part correlation, which is quantified by querying the part semantic dictionary and output in the form of score; TP represents the turnover rate; α1, α2 and α3 are the weights of FR, CB and TP, respectively. In this embodiment, α1=0.2, α2=0.3, and α3=0.5. The trained partition allocation screening model is used to screen the initial part partition set to obtain an optimal part partition set. The optimal partition allocation screening model searches for the optimal solution based on the optimization algorithm to obtain the optimal part partition set that meets the conditions. The optimal part partition set is mapped to the specific storage rack coordinates to obtain the optimal storage layout information. The optimal part partition set is mapped to the specific storage rack through coordinates, which is the optimal storage layout information.

[0063] The part feature classification method for all part feature vectors includes:

[0064] The accessory topology network structure is constructed based on all the accessory feature vectors. In this embodiment, the GNN graph neural network model is used to construct the accessory topology network structure based on all the accessory feature vectors. Each accessory feature vector is taken as a node, and the semantic relationship between the accessory feature vectors is taken as an edge. The feature similarity between any two accessory feature vectors is calculated, and the connection edge between the two accessory feature vectors is weighted by using the feature similarity. The feature similarity is calculated based on the cosine similarity between any two accessory feature vectors, and the edge is weighted by using the similarity, so as to facilitate subsequent reasonable division of the nodes according to the weight of the edge.

[0065] The nodes in the accessory topology network structure are subjected to hierarchical classification processing. A similarity threshold is set, and coarse classification processing is performed based on the feature similarity and the similarity threshold. The accessory node pairs with the feature similarity greater than the similarity threshold are screened to obtain an initial classification node set. The accessory topology network subgraph is constructed based on the initial classification set. The feature similarity is greater, the weight of the edge is greater, and the corresponding accessory node pair has a high correlation degree, thereby forming several accessory clustering areas with strong local similarity. The accessory topology network subgraph is constructed based on the nodes in each local area with strong local similarity in the initial classification node set. Each subgraph represents the connection relationship of the nodes in the local area with strong local similarity. The connection strength between the subgraph accessory nodes in any one accessory topology network subgraph is calculated, and the refinement classification processing is performed based on the connection strength to obtain a fine classification node set. The calculation formula of the connection strength is as follows: wherein CN represents the local aggregation degree of any one accessory node in the accessory topology network subgraph, that is, the connection strength; k0 represents the degree of any one accessory node; e0 represents the number of edges between the nodes connected to the accessory node in the subgraph. The aggregation trend between the nodes is evaluated based on the connection strength. The nodes with a connection strength difference less than a preset connection strength difference threshold are divided into the same fine class, that is, the fine classification node set. The accessory feature classification result is obtained by integrating all the fine classification node sets.

[0066] The path optimization manner of the accessory picking task to be performed includes:

[0067] In this embodiment, the accessory picking task to be performed includes the accessory name, the partition coordinates corresponding to the accessory position, the accessory demand quantity, and the task urgency, and the like.

[0068] The historical path information is collected, the historical path information is path modeled to obtain a working path model, wherein the historical path information represents a historical path selected in a historical record to complete a parts picking task; the historical path is modeled to provide data for subsequent search of a feasible path; a warehouse state space is constructed, and a parts picking task to be performed is path inferred based on the warehouse state space and the working path model to obtain a feasible path, and all feasible paths are integrated to generate a feasible path set, wherein the warehouse state space is used to reflect various attribute data of the warehouse environment, such as information such as shelf number, shelf position, and obstacle area coordinates; constructing the warehouse state space is equivalent to constraining the search range of the path, and by extracting information such as an efficient path and a common blocked area from the working path model, the feasible path is obtained by searching the path in the warehouse state space based on the information of the working path model; a path optimization objective function is constructed, and the feasible path set is path planned based on the path optimization objective function to generate a feasible path score, wherein the calculation formula of the path optimization objective function is: FD = β1 x dis + β2 x ti + β3 x cd; wherein FD represents the feasible path score of any feasible path; dis represents the total length of the feasible path; ti represents the total time consumed by the feasible path; cd represents a judgment of whether the feasible path coincides with an existing path, and if not, the value is 1, and if coincides, the value is 0; β1, β2, and β3 are weights of dis, ti, and cd, respectively, and in this embodiment, β1 = 0.4, β2 = 0.4, and β3 = 0.2, wherein; the path optimization objective function is used to guide the model to search for an optimal path, and the distance, time, and overlap degree are considered at the same time, so as to effectively avoid repeated paths, high time-consuming paths, and low efficiency paths, and improve the work efficiency; a path score threshold is set, and the optimal feasible path is selected based on the feasible path score and the path score threshold, that is, the efficient work optimization path, and the feasible paths with a feasible path score lower than the path score threshold are filtered, and the remaining feasible paths are all high-score and non-repeated paths, which can all be used as efficient work optimization paths for workers to select.

[0069] The preset path guiding unit is deployed, the efficient work optimization path is sent to the path guiding unit, and the path guiding unit is connected with the wearable device, and it should be noted that the wearable device is worn by the worker to reflect the selectable path in real time, thereby reducing the time of manual decision and improving the work efficiency; the efficient work optimization path is visually displayed by using the path guiding unit, a virtual work path scene is generated and sent to the wearable device, in the embodiment, the path guiding unit includes a path receiving module, an instruction generating module, a visualization module and a communication module, the efficient work optimization path is converted into standard navigation instructions by receiving the efficient work optimization path, the corresponding path model is generated based on the navigation instructions by the visualization module, the model is projected and combined with the real scene to obtain the virtual work path scene, the virtual work path scene is sent to the wearable device and directly presented in front of the worker; the work path trajectory is monitored in real time by using the wearable device, and the latest efficient work optimization path is generated based on the work path trajectory in real time by using the path optimization model to adjust the work path trajectory, the data of the work path trajectory monitored in real time by the wearable device is fed back to the path optimization objective function, and the optimal path is calculated based on the path at this time by using the path optimization objective function, so that the path can adapt to the scene change and the work efficiency is maintained.

[0070] The state prediction method of the accessory warehouse entry and exit record comprises:

[0071] The accessory attribute information of the accessory warehouse entry and exit record is extracted, wherein the accessory warehouse entry and exit record comprises the warehouse entry and exit action, time, accessory state and quantity change of the automobile accessory; the accessory attribute information comprises the accessory type, adaptive vehicle type and specification parameter of the corresponding automobile accessory; the accessory warehouse entry and exit record is time-sequenced based on the timestamp of the accessory warehouse entry and exit record, and the warehouse entry and exit time sequence is obtained; the accessory attribute information is data-fused with the warehouse entry and exit time sequence, and the warehouse entry and exit feature time sequence is obtained, in the embodiment, the accessory attribute information is converted into a vector representation, and the accessory attribute vector is obtained; the accessory warehouse entry and exit record corresponding to each timestamp in the warehouse entry and exit time sequence is spliced with the accessory attribute vector of the corresponding accessory at the vector level, and the warehouse entry and exit feature time sequence is obtained.

[0072] A dynamic sliding window is constructed, and the in-and-out warehouse characteristic time series is traversed by using the dynamic sliding window; an accessory state evaluation system is constructed; in this embodiment, an LSTM neural network model is used to capture the state fluctuation trend of the in-and-out warehouse characteristic time series; wherein the accessory state evaluation system comprises an inventory fluctuation index, a replenishment supply cycle index, a turnover rate index and an inventory redundancy index; the state of the in-and-out warehouse characteristic time series segment in each dynamic sliding window is predicted, and a state evaluation score group is output based on the accessory state evaluation system, wherein the in-and-out warehouse characteristic time series segment in each dynamic sliding window is taken as the input data of the LSTM neural network model, the model is used to predict the future inventory change based on the inventory change of the automobile accessory in a certain time segment, and the state evaluation score group is obtained by calculating the inventory fluctuation index, the replenishment supply cycle index, the turnover rate index and the inventory redundancy index in the accessory state evaluation system; wherein the calculation formula of the inventory fluctuation index is: Wherein σST represents the standard deviation of all node inventory values in the in-and-out warehouse characteristic time series segment in any one dynamic sliding window; μST represents the mean of all node inventory values in the in-and-out warehouse characteristic time series segment in any one dynamic sliding window; the calculation formula of the replenishment supply cycle index is: Wherein LT represents the delivery cycle of the supplier of the automobile accessory corresponding to the in-and-out warehouse characteristic time series segment; TS represents the earliest time point of the accessory inventory duration, which is obtained by calculating the ratio of the inventory value corresponding to the starting time stamp of the in-and-out warehouse characteristic time series segment to the consumption amount; the calculation formula of the turnover rate index is: Wherein QST represents the out-of-warehouse quantity in any one dynamic sliding window; ST represents the time span of any one dynamic sliding window; the calculation formula of the inventory redundancy index is: Wherein CC represents the inventory amount at any one time point in the in-and-out warehouse characteristic time series segment; CF represents the predicted inventory demand amount at the time point; ρ represents a minimum constant greater than 0; and the calculation formula of the state evaluation score is: The state evaluation score group is obtained by integrating the state evaluation scores of each dynamic sliding window corresponding to the warehouse-in and warehouse-out feature time sequence segment; a spare part inventory state grading table is designed based on a preset automobile spare part parameter database, the state evaluation score group is compared with the spare part inventory state grading table, and inventory state prediction data of the warehouse-in and warehouse-out feature time sequence segment in a single dynamic sliding window is output, wherein the spare part inventory state grading table includes multiple thresholds, divides the inventory state into several levels, such as inventory shortage, inventory suitable, and inventory warning; the inventory state prediction data includes an inventory state level, a predicted inventory change, and a replenishment strategy, wherein the replenishment strategy is generated based on the demand for different types of automobile spare parts through an inventory intelligent early warning model to predict an inventory difference; the state prediction data of the warehouse-in and warehouse-out feature time sequence segment in each dynamic sliding window is weighted and fused by continuously adjusting the size of the dynamic sliding window, to obtain referenceable inventory information; in this embodiment, different time lengths of features of the warehouse-in and warehouse-out feature time sequence are obtained by continuously adjusting the size of the dynamic sliding window, the state prediction data in different windows is fused, and prediction data with more robustness and trend comprehensiveness is output, wherein the weights of the weighted fusion are set based on existing data or experience.

[0073] The quality detection method of the spare part basic database includes:

[0074] The spare part life cycle data of the spare part basic database is extracted, the spare part life cycle data is feature-fused with the referenceable inventory information of the corresponding automobile spare part, and a quality detection feature vector is obtained; in this embodiment, the spare part life cycle data includes production time, warehouse-in time, maximum use period, and use history time; the spare part life cycle data and the referenceable inventory information are converted into vector representations, and the two types of data are feature-spliced based on a feature level, to obtain the quality detection feature vector; each quality detection feature vector is mapped into a spare part life cycle sequence constructed based on a timestamp, to obtain quality detection trajectory data; the quality detection feature vectors are sequentially arranged according to the timestamps of the spare part life cycle sequence, to construct timestamp-based quality detection trajectory data, wherein the data at each time point includes the spare part feature, the inventory state, the life cycle data, and other information at this moment, and reflects the state change of the spare part within a certain time.

[0075] A parts quality rule database is constructed and used as a constraint condition. In this embodiment, the parts quality rule database is constructed based on existing information, which includes information such as quality anomaly level, environmental impact on parts quality, and solutions for different quality problems. The parts quality rule database is used as a constraint to avoid misidentification of quality problems during quality detection. Based on the parts quality rule database, quality detection trajectory data is subjected to quality detection, and a quality evaluation score and an abnormal state label are output. In this embodiment, a Transformer model is used to model the change trend in the quality detection trajectory data, and a quality evaluation score is calculated to reflect the change trend of the parts quality. In addition, the quality detection trajectory data is subjected to anomaly identification, and quality anomaly problems are extracted. If the quality anomaly problems include the characteristics of the abnormal state of the type in the parts quality rule database, the corresponding abnormal state label is output. The quality evaluation score is essentially a confidence score generated based on the Transformer model. A quality score grading threshold is set and dynamically adjusted. Based on the quality evaluation score and the quality score grading threshold, the quality grade corresponding to each automobile part is determined. The quality score grading threshold corresponding to different types of automobile parts is obtained by querying the parts quality rule database, and the threshold is adjusted based on the type of automobile part corresponding to the quality evaluation score. For each automobile part, the abnormal state label and the quality grade are subjected to risk control, and risk management recommendation instructions for the automobile part are generated in combination with the parts quality rule database. The solution strategy for the quality problem corresponding to the abnormal state label and the corresponding quality grade is obtained by querying the parts quality rule database, and the strategy is converted into a standard system instruction format, i.e., the risk management recommendation instruction. The parts evaluation data is obtained by integrating the quality grade, the abnormal state label, and the risk management recommendation instruction.

[0076] The behavior data is subjected to behavior pattern analysis in the following manner:

[0077] The behavior data is associated with the accessory warehouse entry and exit record to obtain multi-dimensional behavior data, and a worker ID tag corresponding to the behavior data is obtained, and the multi-dimensional behavior data is marked by using the ID tag, wherein the behavior data includes various operation behaviors generated by the worker in work, such as picking scanning and accessory carrying operations; by matching the behavior data and the accessory warehouse entry and exit record based on a time stamp, a corresponding relationship between the operation behavior and the actual business is established; entity extraction is performed on the multi-dimensional behavior data to obtain behavior entities; by structuring the multi-dimensional behavior data, action features such as picking and carrying are extracted, and information such as time, location and ID tag is associated to obtain the behavior entities; a topological graph structure is constructed based on the behavior entities, each behavior entity is taken as a node of a work behavior graph, a logical relationship between any two behavior entities is extracted, and the two behavior entities are semantically connected by using the logical relationship to obtain the work behavior graph, wherein the logical relationship refers to semantic relationships such as time adjacency, operation continuity or path coincidence between the nodes, so that the edges between the nodes in the graph have context semantic information; a behavior mode subgraph of a corresponding worker in the work behavior graph is extracted based on the ID tag, and a behavior path of the corresponding worker is obtained by traversing each node and edge in the behavior mode subgraph, wherein the ID tag is taken as an index, a node set belonging to the same worker and a path connected by the node set are screened out, and a behavior mode subgraph belonging to the worker is constructed; a traversal algorithm is used to traverse the behavior mode subgraph according to time or a logical relationship to obtain one or more directed paths, which are the behavior path of the corresponding worker.

[0078] The historical high evaluation behavior data is collected, and a standard behavior path template is constructed based on the historical high evaluation behavior data. The standard behavior path template is compared with the behavior path, and a behavior deviation feature vector is obtained by calculating the deviation degree. The historical high evaluation behavior data refers to a path with high correct rate, fast execution, and low probability of abnormal situation. The historical high evaluation behavior data is extracted from a preset historical high evaluation behavior data set, and the historical high evaluation behavior data is standardized to obtain the standard behavior path template. The deviation degree is identified by a graph matching algorithm to identify the deviation degree between the standard behavior path template and the behavior path of the staff, and the deviation degree is quantified as a deviation feature vector. The behavior deviation feature vector is evaluated to obtain behavior evaluation information. In this embodiment, the behavior evaluation information is evaluated by using an SVM support vector machine model. The corresponding relationship between the deviation feature vector and the behavior level is identified by training and identifying the deviation feature vector by using sample data, and indexes related to the behavior level of the staff, such as operation efficiency, compliance degree, path deviation degree, and abnormal recognition probability, are output as the behavior evaluation information. The behavior evaluation information is bound to the ID tag of the corresponding staff, and all behavior evaluation information is integrated to obtain behavior evaluation data. The behavior evaluation data is efficiently screened to obtain high evaluation behavior data, and the standard behavior path template is dynamically updated by using the high evaluation behavior data. In this embodiment, a preset high evaluation standard system is used to quantize each index data in the behavior evaluation data and calculate the mean value, and the mean value of each index data in the standard behavior path template is calculated. If the mean value is greater than the mean value of the standard behavior path template, the standard behavior path template is updated based on the behavior evaluation data. Otherwise, it remains unchanged. The accuracy of the behavior evaluation data obtained subsequently is higher and higher by continuously updating the standard behavior path template, and at the same time, the work quality of the staff is improved.

[0079] The warehouse strategy is constructed based on the behavior evaluation data, the accessory evaluation data, the referenceable inventory information, and the efficient work optimization path.

[0080] The behavior evaluation data, the accessory evaluation data, the referenceable inventory information and the efficient work optimization path are respectively subjected to feature extraction to obtain behavior features, accessory quality features, inventory information features and path features, wherein the behavior features refer to features including operation efficiency, compliance degree and behavior path corresponding to operation; the accessory quality features refer to features including accessory quality classification, abnormality identification label and life cycle information; the inventory information features refer to features including warehouse-in and warehouse-out records, inventory quantity and change trend and predicted inventory change amount; and the path features are the efficient work optimization path. The behavior features, the accessory quality features, the inventory information features and the path features are subjected to feature fusion to obtain a multi-dimensional joint feature vector. The behavior features, the accessory quality features, the inventory information features and the path features are subjected to vector splicing based on semantic correlation to obtain a multi-dimensional joint feature vector. The vector embodies description information of a complete task execution scene. Sub-dimensions of the multi-dimensional joint feature vector are respectively provided with judgment indexes, and the judgment indexes are matched with a preset strategy rule database. A multi-dimensional joint feature vector corresponding to all judgment indexes capable of matching the strategy rule database is taken as a warehouse strategy, wherein the judgment indexes include data such as behavior level, inventory risk level, accessory quality level and path efficiency, which can judge the advantages and disadvantages of corresponding feature information in executing work. The strategy rule database includes various condition combinations corresponding to complete task execution scenes available in historical records. If the judgment indexes satisfy the strategy triggering condition of the condition combination, the multi-dimensional joint feature vector is determined as the warehouse strategy, and the vector is converted into a strategy instruction recognizable by the system.

[0081] The embodiment realizes an intelligent automobile accessory warehouse management method and system based on the Internet of Things from aspects of data collection, perception fusion, partition optimization, path planning, inventory prediction, quality monitoring, behavior analysis and strategy generation, and significantly improves the overall operation efficiency and decision-making ability of the warehouse system. Compared with the traditional mode relying on manual experience and static strategy, the intelligent automobile accessory warehouse management method and system based on the Internet of Things can fully fuse accessory data, Internet of Things perception data and behavior data, establish semantic mapping and state perception of multi-source information, improve accessory storage layout and work efficiency through intelligent partition and path optimization, enhance risk prevention and control capability through prediction and detection of inventory state and accessory quality, and indirectly optimize the work level of workers by analyzing behavior data of the workers. Therefore, the intelligent automobile accessory warehouse management method and system based on the Internet of Things has technical advantages such as high intelligence level, strong robustness and good self-adjusting capability, and can be widely applied to warehouse management scenes in the fields of intelligent manufacturing and automobile service, and has good practical value and promotion prospect.

[0082] Embodiment 2

[0083] Please refer to Figure 2As shown, the embodiments not described in detail see example 1, provide a kind of based on Internet of Things's car accessory intelligent warehouse management system, comprising:

[0084] Data acquisition module, for collecting automobile parts attribute data and performing accessory labeling, obtaining labeled accessory dataset;Standard accessory dataset is cleaned to obtain accurate accessory dataset;Based on accurate accessory dataset, accessory basic database is constructed;

[0085] Accessory partition module, for collecting original perception data and associating accessory basic database, obtaining multi-source storage dataset;Multi-source storage dataset is partitioned, and optimal warehouse layout information is generated;

[0086] Path optimization module, for obtaining accessory real-time position information, based on optimal warehouse layout information and accessory real-time position information, the accessory picking task to be executed is path optimized, and efficient work optimization path is output;

[0087] Inventory prediction module, for real-time acquisition of accessory warehouse record, state prediction is carried out on accessory warehouse record, and referenceable inventory information is output;

[0088] Quality detection module, for quality detection of accessory basic database based on referenceable inventory information, and accessory evaluation data is output;

[0089] Behavior analysis module, for collecting behavior data of staff, behavior pattern analysis is carried out on behavior data, and behavior evaluation data is generated;

[0090] Strategy generation module, for constructing warehouse strategy based on behavior evaluation data, accessory evaluation data, referenceable inventory information and efficient work optimization path, and sending warehouse strategy to preset automobile accessory management system;Each module is connected by wired and / or wireless mode.

[0091] The above only for the preferred embodiment of the application, and not for limiting the application, although the application is described in detail with reference to the foregoing embodiments, for those skilled in the art, still can modify the technical scheme recorded in the foregoing each embodiment, or equivalent replacement to part of technical features.It is any modification, equivalent replacement, improvement, etc., which is made within the spirit and principles of the application, should be included in the protection scope of the application.

[0092] It should be noted that, in the present document, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can also include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element preceded by "comprises... a" does not, without more constraints, foreclose the existence of additional identical elements in the process, method, article, or apparatus that comprises the recited element.

[0093] In the description of the present application, it should be understood that the terms "first", "second" and the like are used to distinguish one element from another, and are not necessarily used to describe a relative importance level or order.

[0094] In the description of the present application, the meaning of "a plurality of" is two or more, unless otherwise specified.

[0095] In the description of the present application, the meaning of "a plurality of" is two or more, unless otherwise specified.

[0096] In the description of the present application, the description of the terms "one embodiment", "some embodiments", "an example", "a specific example", or "some examples" and the like means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the present specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0097] For the formulas in the present specification, the values are calculated by de-dimensioning, the formulas are obtained by software simulation based on a large amount of data to obtain a formula closest to the real situation, and the preset parameters and threshold values in the formulas are set by a person skilled in the art according to the actual situation.

[0098] Although the embodiments of the present application have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, replacements and variations can be made to the embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the claims and their equivalents.

Claims

1. An intelligent storage management method for automobile parts based on the Internet of Things, characterized in that: include: S1. Collect auto parts attribute data and label the parts to obtain a labeled parts dataset; Perform data cleaning on the standard parts dataset to obtain the precise parts dataset; build a basic parts database based on the precise parts dataset; S2. Collect raw sensory data and associate it with the basic parts database to obtain a multi-source storage dataset; partition the multi-source storage dataset into parts to generate optimal storage layout information; S3. Obtain real-time parts location information, optimize the path for parts picking tasks based on the optimal warehouse layout and real-time parts location information, and output an optimized and efficient work path. S4. Real-time collection of parts in and out of the warehouse, status prediction of parts in and out of the warehouse, and output of reference inventory information; S5. Perform quality inspection on the basic parts database based on available inventory information and output parts evaluation data; S6. Collect behavioral data of staff members, analyze behavioral patterns of the behavioral data, and generate behavioral assessment data; S7. Build a warehousing strategy based on the behavior evaluation data, parts evaluation data, reference inventory information, and efficient work optimization path, and send the warehousing strategy to the preset automotive parts management system.

2. The method for intelligent storage management of automobile parts based on the Internet of Things according to claim 1, characterized in that: The method of collecting the original perception data and associating it with the basic accessory database includes: The original perception data is collected and features are extracted from the original perception data to obtain original perception features; an accessories semantic dictionary is constructed based on a preset automobile accessories parameter database; accessory tags are extracted from the accessories basic database, the accessories basic database is grouped based on the accessory tags, and a unique correspondence between each accessory tag and the corresponding group of accessories basic data is established; the original perception features are semantically matched with each group of accessories basic data and accessory tags based on the accessories semantic dictionary to obtain a multi-source storage data set.

3. The method for intelligent storage management of automobile parts based on the Internet of Things according to claim 2, characterized in that: The method of performing accessory partitioning on a multi-source storage data set includes: Feature extraction is performed on multi-source storage datasets to obtain structured semantic features. A feature vector space is constructed and all structured semantic features are mapped to the feature vector space to obtain accessory feature vectors. All accessory feature vectors are classified to obtain accessory feature classification results. Based on the accessory feature classification results, accessories of the same category are divided into the same storage area to obtain the initial accessory partition set. Extract the accessories warehouse layout data in real time, and build a partition allocation screening model based on the accessories warehouse layout data; construct the optimal screening function, and use the optimal screening function as the loss function of the partition allocation screening model; use the trained partition allocation screening model to partition the initial accessories partition set to obtain the optimal accessories partition set; map the optimal accessories partition set to specific storage shelf coordinates to obtain the optimal warehouse layout information.

4. The method for intelligent storage management of automobile parts based on the Internet of Things according to claim 3, characterized in that: The method of classifying all accessory feature vectors includes: Construct an accessory topology network structure based on all accessory feature vectors; calculate the feature similarity of any two accessory feature vectors, and use the feature similarity to weight the connection edge between the two accessory feature vectors; All accessory nodes in the accessory topology network structure are hierarchically classified, a similarity threshold is set, and coarse classification is performed based on feature similarity and similarity threshold. Accessory node pairs with feature similarity greater than the similarity threshold are screened to obtain an initial classification node set, and an accessory topology network subgraph is constructed based on the initial classification set; the connection strength between subgraph accessory nodes in any accessory topology network subgraph is calculated, and refined classification is performed based on the connection strength to obtain a fine classification node set; all fine classification node sets are integrated to obtain the accessory feature classification result.

5. The method for intelligent storage management of automobile parts based on the Internet of Things according to claim 4, characterized in that: The method of optimizing the path of the parts picking task to be performed includes: Collect historical path information, perform path modeling on the historical path information, and obtain a work path model; construct a warehouse state space, and perform path reasoning on the parts picking task to be executed based on the warehouse state space and the work path model to obtain a feasible path, and integrate all feasible paths to generate a feasible path set; construct a path optimization objective function, and perform path planning on the feasible path set based on the path optimization objective function to generate a feasible path score; set a path score threshold, and screen the optimal feasible path based on the feasible path score and the path score threshold, which is the efficient work optimization path; Deploy a preset path guidance unit, send the efficient work optimization path to the path guidance unit, and establish a connection between the path guidance unit and the wearable device; use the path guidance unit to visualize the efficient work optimization path, generate a virtual work path scene and send it to the wearable device; use the wearable device to monitor the work path trajectory in real time, and use the path optimization model to generate the latest efficient work optimization path in real time based on the work path trajectory to adjust the work path trajectory.

6. The method for intelligent storage management of automobile parts based on the Internet of Things according to claim 5, characterized in that: The method for predicting the status of the parts in and out of the warehouse includes: Extract the attribute information of auto parts from the parts in and out records; sort the parts in and out records by time based on their timestamps to obtain an in and out time series; fuse the attribute information with the in and out time series to obtain an in and out feature time series; A dynamic sliding window is constructed and used to traverse the time series of in-and-out feature. A parts status assessment system is constructed. Status prediction is performed on the in-and-out feature time series segments within each dynamic sliding window, and status assessment score groups are output based on the parts status assessment system. A parts inventory status grading table is designed based on a preset automotive parts parameter database. The status assessment score groups are compared with the parts inventory status grading table, and inventory status prediction data for the in-and-out feature time series segments within a single dynamic sliding window is output. By continuously adjusting the size of the dynamic sliding window, the status prediction data of the in-and-out feature time series segments within each dynamic sliding window are weighted and fused to obtain reference inventory information.

7. The method for intelligent storage management of automobile parts based on the Internet of Things according to claim 6, characterized in that: The method of performing quality inspection on the basic parts database includes: Extract the accessory lifecycle data from the basic accessory database and fuse the feature of the accessory lifecycle data with the reference inventory information of the corresponding auto parts to obtain the quality inspection feature vector. Map each quality inspection feature vector to the accessory lifecycle sequence constructed based on the timestamp to obtain the quality inspection trajectory data. Construct an accessories quality rule database and use it as a constraint; perform quality inspection on quality inspection trajectory data based on the accessories quality rule database, and output quality assessment scores and abnormal status labels; set and dynamically adjust quality score grading thresholds, and determine the quality grade corresponding to each auto part based on the quality assessment score and quality score grading thresholds; perform risk management and control on the abnormal status label and quality grade corresponding to each auto part, and generate risk management recommendation instructions for the auto part in combination with the accessories quality rule database; integrate quality grades, abnormal status labels, and risk management recommendation instructions to obtain accessories evaluation data.

8. The method for intelligent storage management of automobile parts based on the Internet of Things according to claim 7, characterized in that: The method of analyzing the behavior pattern of the behavior data includes: The behavioral data is associated with the parts in and out records to obtain multidimensional behavioral data, and the staff ID tags corresponding to the behavioral data are obtained, and the multidimensional behavioral data are marked with the ID tags; entity extraction is performed on the multidimensional behavioral data to obtain behavioral entities; a topological graph structure is constructed based on the behavioral entities, and each behavioral entity is used as a node in the work behavior graph. The logical relationship between any two behavioral entities is extracted, and the two behavioral entities are semantically connected using the logical relationship to obtain the work behavior graph; the behavioral pattern subgraph corresponding to the staff member in the work behavior graph is extracted based on the ID tag, and the behavioral path of the corresponding staff member is obtained by traversing each node and edge in the behavioral pattern subgraph; Collect historical highly-rated behavior data, and build a standard behavior path template based on the historical highly-rated behavior data. Compare the standard behavior path template with the behavior path, and calculate the degree of deviation to obtain the behavior deviation feature vector; perform behavior evaluation on the behavior deviation feature vector to obtain behavior evaluation information; bind the behavior evaluation information with the ID tag of the corresponding staff member, and integrate all behavior evaluation information to obtain behavior evaluation data; at the same time, perform efficient behavior screening on the behavior evaluation data to obtain highly-rated behavior data, and use the highly-rated behavior data to dynamically update the standard behavior path template.

9. The method for intelligent storage management of automobile parts based on the Internet of Things according to claim 8, characterized in that: The method of constructing a warehousing strategy based on behavior evaluation data, parts evaluation data, reference inventory information, and efficient work optimization path includes: Feature extraction is performed on behavior evaluation data, parts evaluation data, reference inventory information and efficient work optimization path respectively to obtain behavior features, parts quality features, inventory information features and path features; feature fusion is performed on behavior features, parts quality features, inventory information features and path features to obtain a multidimensional joint feature vector; judgment indicators are set for the sub-dimensions of the multidimensional joint feature vector respectively, and the judgment indicators are matched with the preset strategy rule database, and the multidimensional joint feature vector in which all judgment indicators can be matched with the corresponding one in the strategy rule database is used as a warehousing strategy.

10. An intelligent storage management system for automobile parts based on the Internet of Things, which is used to implement an intelligent storage management method for automobile parts based on the Internet of Things according to any one of claims 1 to 9, characterized in that: include: The data acquisition module is used to collect attribute data of automobile parts and label the parts to obtain a labeled parts data set; Perform data cleaning on the standard parts dataset to obtain the precise parts dataset; build a basic parts database based on the precise parts dataset; The parts partitioning module is used to collect raw perception data and associate it with the basic parts database to obtain a multi-source storage data set; it partitions the multi-source storage data set into parts partitions to generate optimal storage layout information; The path optimization module is used to obtain the real-time location information of parts, optimize the path of the parts picking tasks to be executed based on the optimal warehouse layout information and the real-time location information of parts, and output the efficient work optimized path; The inventory forecast module is used to collect the parts in and out of the warehouse in real time, predict the status of the parts in and out of the warehouse, and output reference inventory information; The quality inspection module is used to perform quality inspection on the basic database of accessories based on reference inventory information and output accessory evaluation data; Behavior analysis module, used to collect behavioral data of staff members, analyze behavioral patterns of the behavioral data, and generate behavioral evaluation data; The strategy generation module is used to build a warehousing strategy based on behavior evaluation data, parts evaluation data, reference inventory information and efficient work optimization paths, and send the warehousing strategy to the preset automotive parts management system; each module is connected via wired and / or wireless means.

Images